The present invention relates to a system and method for differential pulse code modulation (DPCM) transmission, and more particularly to a system and method for adaptive DPCM (ADPCM) transmission to achieve band compression of speech or like signals.
For speech or like signals which are closely correlated to one another, the amplitude of a sample can be predicted on the basis of previous samples. A DPCM system achieves band compression by utilizing this feature of speech or like signals. The DPCM system is equipped with a predictor whose prediction coefficients are determined by the average speech spectrum. However, since actual speech signals vary with time, the optimum values of the coefficients also vary with time. As a result, the DPCM system having fixed prediction coefficients cannot be very precise. For the purpose of obviating this disadvantage of the DPCM system, there has been developed an ADPCM system in which prediction coefficients are adaptively corrected in accordance with speech signals varying with time. For details of the ADPCM system, reference is made to U.S. Pat. No. 4,317,208 and to a paper by David L. Cohn et al., entitled "The Residual Encoder--An Improved ADPCM System for Speech Digitalization", IEEE Transactions on Communications, Vol. COM-23, No. 9, pp. 935-941.
However, because the prediction coefficients are constantly corrected in the ADPCM system, there arises the problem of predictor stability, which problem does not occur when the prediction coefficients are fixed. Thus, if the coefficients of the adaptive predictor happens to be corrected to such a value as causes the oscillation of the predictor, the oscillation results in a predicted value which is unrelated to the input signal and thereby prevents precise prediction.